Lilian’s suggestions

• Find a Hugging face model on fluency to use

• Use ML split testing/training dataset

• training, testing, validation: K fold

• Try PCA for readability

• class imbalance, data augmenatation

• grammarly high level solutions

• talk to people who do NLP

• Feed BERT: fine tuning

• ML fluency model :Speaker Fluency Level Classification Using Machine Learning Techniques

• How to know if the machine is accurate?

  • Sample and score yourself: I score this way, and machine score by that way

• Paste all the texts in grammarly and compare grammarly score with my scores

• I can probably do a few clusters for all my fluency related score to see if things related with each other

• Rather than doing accurate prediction, can we investigate do group difference on fluency exists among different grade groups

• Interactive website: input text with fluency score

  https://unfair-machine.vuluong.me/blinding/

• Baysien inference

• Ask Diamond on causal inference (matching)

• Discussion session: what’s the difference between other research results and mine?

• High scores vs low scores: tf-idf difference

• Ask Rebecca on writing a research paper